Methods Lab
Methods that make assumptions visible.
Local tools for reviewing data and interpreting study quantities with explicit assumptions.
Available now
Effect Translator
Translate group risks, RR, OR, RD or a compatible HR into fixed-time absolute effects with explicit assumptions and limitations.
Data quality
Data Check
Review CSV missing values, types, duplicates and declared rules locally in your browser. Preserve the original file and download a descriptive report.
Simulations
Survival Lab
Explore synthetic survival, follow-up and independent censoring with Kaplan–Meier curves, uncertainty and reproducible exports.
Simulations
Trial Lab
Simulate fixed-size binary trials: risk differences, Newcombe intervals, Fisher tests and Monte Carlo uncertainty across up to five scenarios.
Prospective planning
Trial Planner
Plan sample size, power and detectable risk differences for two independent binary groups. Explore assumptions and export reproducible calculations.
Local processing
Your inputs stay in your browser.
All five tools process their working inputs in browser memory. Effect Translator uses aggregate quantities; Trial Planner uses prospective aggregate assumptions; Survival Lab and Trial Lab generate synthetic data. Data Check reads a local CSV without uploading it; a downloaded report may contain its values. Use synthetic or public, non-identifiable data. Do not enter patient-level, confidential or identifiable data.
Method and limitations
Assumptions and limitations of every tool.
These tools support descriptive checks and methodological exploration. Their outputs depend on the selected inputs and assumptions; they do not establish causality, validate a clinical protocol or replace statistical review.
Effect Translator
Absolute-effect translations depend on a relevant comparator risk, a compatible population and the same outcome and time horizon. HR is not RR: its translation requires proportional hazards and time to first event. Competing-risk HRs, recurrent events, time-varying HRs, non-proportional hazards and RMST are unsupported.
Read Effect Translator assumptions
Data Check
Checks describe the selected file according to confirmed types and declared rules. Repeated identifiers may be valid visits; IQR outlier candidates require context. It does not clean or impute data, infer why data are missing, fit models or certify clinical data quality. Limits: 5 MiB, 20,000 rows, 100 columns and 500,000 cells.
Read Data Check scope and limits
Survival Lab
Simulations assume two synthetic groups, exponential events, constant proportional hazards, simultaneous entry and independent dropout. Latent dropout probability is not the observed censored fraction. Kaplan–Meier intervals are pointwise 95% log–log intervals, not simultaneous bands; intervals at survival 0 or 1 are unavailable. No Cox model, log-rank test, p-values, power, competing risks or real-data import are included.
Read Survival Lab scope and limits
Trial Lab
Simulations use independent binary groups with sample sizes fixed in advance. Newcombe–Wilson interval coverage is approximate and may fall below 95%; the two-sided Fisher test may be conservative. The interval and test are distinct procedures. Estimated power or null rejection depends on the scenario and has Monte Carlo uncertainty. No dropout, censoring, non-adherence, clustered groups or multiple outcomes are included.
Read Trial Lab scope and limits
Trial Planner
Normal pooled-Z planning for two independent binary groups, one final test and a fixed horizon. Common loss inflation uses integer allocation blocks. Sensitivity is conditional; detectable RD uses a 0.01 percentage-point grid. Approximate power can differ from finite-sample rejection and is not probability of benefit. This is a different method from Trial Lab Fisher simulations.